Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by motor and non-motor symptoms1. The core pathophysiology of PD involves the loss of dopaminergic neurons in the substantia nigra pars compacta (SNpc) and disruption of basal ganglia circuits2,3. However, PD-related pathological changes are not limited to the dopamine system and basal ganglia but also affect multiple neuroanatomical regions, including the cerebellum4. Recent animal studies have shown that the cerebellum directly, rapidly, and effectively modulates dopamine levels in the basal ganglia by regulating the dopaminergic activity in the substantia nigra, and conveys information related to movement initiation5. Additionally, previous functional neuroimaging studies have found a broad and close connection between the cerebellum and PD6,7. These findings suggest that the cerebellum plays a critical role in the pathophysiology and pathogenesis of PD.

Building on these insights, FDG-PET studies further indicate that, beyond its primary role in motor control, the cerebellum also plays a significant role in cognitive functions in PD patients8. Compared to age-matched healthy controls, PD patients exhibit a unique metabolic pattern, known as the PD-related pattern (PDRP), one of the characteristics of which is hypermetabolism in the cerebellum9. Similarly, the PD-related cognitive pattern (PDCP), associated with cognitive impairment in PD, also shows increased metabolism in the cerebellum10,11. However, the precise role of this hypermetabolism in PD remains unclear. Different cerebellar regions may be associated with distinct clinical symptoms, and it is still uncertain whether the observed hypermetabolism reflects underlying pathological changes or a compensatory mechanism. These questions warrant further investigation and may provide new insights into the understanding of PD.

This study aims to analyze the patterns of cerebellar glucose metabolism in PD patients and explore the potential mechanisms affecting motor and cognitive functions using FDG-PET imaging. Additionally, subgroup analyses were conducted to evaluate the consistency of these metabolic patterns among PD patients across different demographic and clinical profiles.

Results

Demographic analysis

A total of 104 patients diagnosed with PD were enrolled in the study, including 33 in the tremor-dominant (TD) group, 61 in the postural instability gait difficulty (PIGD) group, and 10 in the intermediate PD group. In addition, the PD-No Cognitive Impairment (PD-NC) group consisted of 26 patients, the PD-Mild Cognitive Impairment (PD-MCI) group had 40 patients, and the PD Dementia (PDD) group included 38 patients. It is important to note that the number of patients in the Intermediate PD group was relatively small, and their clinical features overlapped with those of the other two groups. Therefore, this group was not included in further subtype analyses but was retained in analyses involving the entire patient cohort. Demographic and clinical characteristics of each group are shown in Table 1.

Table 1 Demographic and clinical data of patients
Full size table

For motor subtypes, the analysis of variance (ANOVA) revealed significant differences between the TD and PIGD groups in disease duration, education level, levodopa equivalent daily dose (LEDD), Movement Disorder Society-Unified Parkinson’s Disease Rating Scale-III (MDS-UPDRS-III) scores, and Montreal Cognitive Assessment (MoCA) scores. Specifically, the TD group had significantly shorter disease duration (p = 0.006), higher education levels (p = 0.015), and lower LEDD (p = 0.005) compared to the PIGD group. Additionally, the MDS-UPDRS-III scores indicated that motor symptoms were more severe in the PIGD group than in the TD group (p = 0.008). The MoCA scores further demonstrated that cognitive function was significantly lower in the PIGD group compared to the TD group (p = 0.004).

In terms of cognitive state classification, ANOVA results revealed significant differences among the three groups in terms of age, disease duration, LEDD, education level, and MDS-UPDRS-III scores. Specifically, patients in the PD-NC group were significantly younger than those in the PDD group (p = 0.006), The disease duration in the PD-NC group was significantly shorter than in the other two groups (p < 0.001). The LEDD in the PDD group was significantly higher than that in the other two groups (PD-MCI, p < 0.002; PD-NC, p < 0.001). The education level in the PD-NC group was notably higher than that of the other two groups (p < 0.001). The MDS-UPDRS-III scores indicated a significant difference in the severity of motor symptoms across the three groups (p < 0.001), with motor symptoms worsening as cognitive function declined.

Correlations between MDS-UPDRS-III scores and cerebellar metabolism

We applied a significance threshold of p < 0.005 at the uncorrected peak level, with a minimum cluster size of 30 voxels.

After controlling for covariates such as age, gender, disease duration, LEDD, and MOCA scores, MDS-UPDRS-III scores were positively correlated with glucose metabolism in the left cerebellar lobules IV and V. For details, refer to Fig. 1 and Supplementary Table 1.

Fig. 1: Cerebellar glucose metabolism correlations with MoCA and UPDRS-III scores.
figure 1

Red: Regions positively correlated with UPDRS-III scores, located in the left cerebellar lobules IV and V; Blue: Regions negatively correlated with MoCA scores, located in the bilateral cerebellar lobules VI and VIII, as well as the bilateral dentate nuclei. Significance threshold: For the UPDRS-III correlation analysis: p < 0.005 (uncorrected), with voxels > 30; For the MoCA correlation analysis: p < 0.005 (uncorrected, peak level), with FWE correction at the cluster level. At this threshold, no regions were found to be positively correlated with MoCA or negatively correlated with UPDRS-III.

Full size image

Violin plot and regression curve suggest (Fig. 2): glucose metabolism in cerebellar lobules IV and V was positively correlated with UPDRS-III scores (p < 0.001, r = 0.382). However, there were no significant differences in glucose metabolism values of cerebellar lobules IV and V between the two groups (p = 0.309). The results of ANOVA and analysis of covariance (ANCOVA) are consistent. For details, refer to Fig. 2 and Supplementary Table 2.

Fig. 2: Cerebellar glucose metabolism differences in Parkinson’s disease subgroups and their correlation with cognitive and motor functions.
figure 2

Violin plots (left) show differences in glucose metabolism in three significant brain regions across different PD subgroups. Regression curves (right) demonstrate the correlations between glucose metabolism and cognitive function (MoCA scores) or motor function (UPDRS-III scores). A Cerebellar lobule VI and MoCA scores: Violin plot: ANOVA revealed significant differences among the three groups (p < 0.001). Glucose metabolism in the PDD group was significantly higher than in the PD-NC group (p < 0.001) and the PD-MCI group (p = 0.005). The PD-MCI group also showed higher metabolism than the PD-NC group (p = 0.038).Regression curve: Glucose metabolism was significantly negatively correlated with MoCA scores (p < 0.001, r = −0.451). B Cerebellar lobule VIII and MoCA scores: Violin plot: ANOVA showed significant group differences (p < 0.001). The PDD group had higher glucose metabolism than the PD-NC group (p < 0.001) and the PD-MCI group (p = 0.013), and the PD-MCI group also had higher metabolism than the PD-NC group (p = 0.032). Regression curve: A significant negative correlation between glucose metabolism and MoCA scores was observed (p < 0.001, r = −0.428). C Cerebellar lobules IV and V and UPDRS-III scores: Violin plot: No significant differences were found between the TD and PIGD groups (p = 0.309). Regreon curve: Glucose metabolism was positively correlated with UPDRS-III scores (p < 0.001, r = 0.382). *p < 0.05, **p < 0.01. For comparisons among three groups, ANOVA followed by Bonferroni correction was used. For comparisons between the two groups, the Mann–Whitney test was applied.

Full size image

Correlation between MoCA scores and cerebellar metabolism

We applied a significance threshold of p < 0.005 at the uncorrected peak level and family-wise error (FWE) correction at the cluster level.

After controlling for covariates such as age, gender, disease duration, education level, LEDD, and MDS-UPDRS-III scores, MoCA scores were negatively correlated with bilateral dentate nucleus, cerebellar lobes VI, and cerebellar lobes VIII in all PD patients. For details, refer to Fig. 1 and Supplementary Table 1.

Violin plot and regression curve suggest (Fig. 2): there was a significant negative correlation between glucose metabolism in cerebellar lobule VI and MoCA scores (p < 0.001, r = -0.451). Group comparisons revealed that the PDD group had higher glucose metabolism values in lobule VI than both the PD-NC group (p < 0.001) and the PD-MCI group (p = 0.005), while the PD-MCI group had higher values than the PD-NC group (p = 0.038). Similarly, glucose metabolism in cerebellar lobule VIII was negatively correlated with MoCA scores (p < 0.001, r = −0.428). Group comparisons showed that the PDD group had higher glucose metabolism values in lobule VIII than both the PD-NC group (p < 0.001) and the PD-MCI group (p = 0.013), and the PD-MCI group had higher values than the PD-NC group (p = 0.032). The results of ANOVA and ANCOVA are consistent. For details, refer to Fig. 2 and Supplementary Table 2.

Regression model construction and subgroup analysis

The results of the regression models are shown in Table 2.

Table 2 Regression models
Full size table

In the motor domain, the regression model results indicate that MDS-UPDRS-III score and sex significantly affect glucose metabolism in cerebellar lobules IV and V. Specifically, MDS-UPDRS-III score is positively associated with glucose metabolism in these regions (β = 0.166, p = 0.001). Gender also has a significant effect, with females showing lower glucose metabolism compared to males (β = -3.991, p = 0.003). Other factors, including age (p = 0.640), disease duration (p = 0.086), LEDD (p = 0.081), and MoCA score (p = 0.663), do not significantly influence glucose metabolism in cerebellar lobules IV and V. The overall regression model explained 21.6% of the variance in cerebellar lobules IV and V glucose metabolism, with an Adjusted R² of 0.216. The model was statistically significant, with an F-statistic of 5.735 (p < 0.001).

In the cognitive domain, the regression model results indicate that MoCA score and sex significantly affect glucose metabolism in both cerebellar lobules VI and VIII. Specifically, MoCA score is negatively associated with glucose metabolism in these regions (VI: β = -0.457, p < 0.001; VIII: β = −0.453, p < 0.001). Gender also has a significant effect, with females showing lower glucose metabolism compared to males (VI: β = −2.606, p = 0.004; VIII: β = -3.225, p = 0.001). Age is positively associated with cerebellar metabolism in cerebellar lobules VI (VI: β = 0.217, p = 0.004; VIII: β = 0.141, p = 0.073). However, other factors, including disease duration (VI: p = 0.673; VIII: p = 0.760), LEDD (VI: p = 0.248; VIII: p = 0.451), education level (VI: p = 0.187; VIII: p = 0.712), and MDS-UPDRS-III scores (VI: p = 0.580; VIII: p = 0.400), do not significantly influence glucose metabolism in either cerebellar lobule. The overall regression models explained 32.3% (VI) and 27.3% (VIII) of the variance in cerebellar lobule glucose metabolism, with Adjusted R² values of 0.323 (VI) and 0.273 (VIII). Both models were statistically significant, with F-statistics of 8.022 (VI, p < 0.001) and 6.520 (VIII, p < 0.001).

The results of the subgroup analysis are shown in Table 3.

Table 3 Subgroup analysis of glucose metabolism in meaningful regions of the cerebellum
Full size table

The relationship between MDS-UPDRS-III scores and glucose metabolism varied across subgroups. Significant positive correlations were observed in the LEDD ≤ 300, disease duration ≤ 6 years, ≤ 67 years age group, and male patients, while no significant correlation was found in the LEDD > 300, disease duration > 6 years, >67 years age group or female patients.

The relationship between MOCA scores and glucose metabolism varied across subgroups. Significant negative correlations were observed in both cerebellar lobules VI and VIII in the male and female subgroups, the age ≤ 67 years and >67 years subgroups, the LEDD ≤ 300 and LEDD > 300 subgroups, the disease duration >6 years subgroup, and the education ≤12 years subgroup. On the other hand, in the disease duration ≤6 years subgroup, significant negative correlations were only observed in cerebellar lobule VI. No significant correlations were found in patients with >12 years of education.

Patterns of metabolic correlations between cerebellar clusters and supratentorial structures

We applied a significance threshold of p < 0.001 at the uncorrected peak level, along with FWE correction at the cluster level for cerebellar lobules IV and V (PIGD group), cerebellar lobule VI, and cerebellar lobule VIII. Additionally, for cerebellar lobules IV and V (TD group), a significance threshold of p < 0.005 was applied at the uncorrected peak level, with a minimum cluster size of 40 voxels.

In the motor domain, in patients of the TD group, glucose metabolism in cerebellar lobules IV and V was positively correlated with glucose metabolism in the thalamus, as well as the parietal, and occipital cortices. However, in the PIGD group, this positive correlation with the thalamus was absent, showing positive correlations with the metabolic activity in the frontal cortex. In the cognitive domain, glucose metabolism in cerebellar lobules VI and VIII was positively correlated with glucose metabolism in the frontal, temporal, parietal, and occipital cortices. For further details, please refer to Fig. 3 and Supplementary Table 3.

Fig. 3: Correlation patterns between different cerebellar regions and supratentorial structures.
figure 3

Red indicates positive correlations. Significance thresholds: A, B, D p < 0.001 (uncorrected, peak level), FWE correction at cluster level; Cp < 0.005 (uncorrected), voxel size > 40. A Cerebellar lobule VI: positively correlated with the metabolic activity of the frontal, parietal, temporal, and occipital cortices. B Cerebellar lobule VIII: positively correlated with the metabolic activity of the frontal, parietal, temporal, and occipital cortices. C Cerebellar lobules IV and V (TD group): positively correlated with the thalamus, temporal, and occipital cortices. D Cerebellar lobules IV and V (PIGD group): positively correlated with the frontal cortex. No significant brain regions showed negative correlations.

Full size image

Discussion

In this retrospective study, we observed several key findings. First, in the motor domain, glucose metabolism in the left cerebellar lobules IV and V is positively correlated with motor function scores (assessed by MDS-UPDRS-III) in PD patients. Second, in the cognitive domain, glucose metabolism in bilateral cerebellar lobules VI, VIII, and the bilateral dentate nucleus is negatively correlated with global cognitive function scores (assessed by MoCA) in PD patients. Third, in the subgroup analysis, a significant positive correlation was found between MDS-UPDRS-III scores and glucose metabolism in male patients, younger individuals, those with shorter disease duration, and those on a lower LEDD. Similarly, the relationship between MOCA scores and glucose metabolism showed significant negative correlations in patients with longer education years, as well as consistently across subgroups based on sex, age, and LEDD. Lastly, in PD patients, the glucose metabolism in the aforementioned cerebellar cortex is positively correlated with supratentorial structures, especially the cerebral cortex, in both motor and cognitive domains. Notably, the TD group showed a positive correlation with the thalamus.

It is well-established that, in healthy individuals, the cerebellum plays a crucial role in motor function. Similarly, studies in PD patient populations have shown that cerebellar atrophy12,13, hypermetabolism8, and functional connectivity6,7,14 within the cerebellum are associated with the severity of motor symptoms. However, exploration at the molecular level remains limited. Our findings show that glucose metabolism in cerebellar lobules IV and V is positively correlated with the severity of motor symptoms in PD patients, indicating that increased glucose metabolism in these areas may play a compensatory role in motor function for PD patients. These lobules, including parts of the vermis, are traditionally regarded as motor-related regions, which aligns with our observations. A previous principal component analysis study conducted on patients undergoing preoperative evaluation for deep brain stimulation (DBS) identified that increased glucose metabolism in the cerebellar lobules VI and VIII, as well as parts of the vermis, correlates with worsening motor symptoms8. The significant regions identified in our study differ from those in the previous study, possibly due to the relatively mild motor symptoms of patients in our study and the differences in methods used to assess motor symptoms between the two studies.

Both structural MRI (sMRI) and functional MRI (fMRI) studies have highlighted the cerebellum as an important region associated with tremors in PD patients7,13. However, the role of the cerebellum in axial symptoms, while not as widely recognized as in tremors, has also been suggested in some studies. Specifically, research has found cerebellar network activation during motor tasks in PD-PIGD patients15,16. In our study, comparing glucose metabolism in key cerebellar regions between the TD and PIGD groups revealed no significant differences. This suggests that the cerebellum may play a common role in PD patients, helping to maintain motor function. Further correlation analysis of significant cerebellar regions with supratentorial structures revealed that cerebellar glucose metabolism in TD patients was not only associated with the cerebral cortex but also closely linked with the thalamus. In the PIGD group, however, no correlation with the thalamus was found, while key voxels in the dorsolateral prefrontal cortex were significantly increased. Given that tremor is the primary clinical manifestation in the TD group, this finding further supports the role of the cerebellar-thalamo-cortical pathway in tremor-related circuits17 and suggests that cerebellar involvement in PIGD patients may bypass the thalamus18, indicating the presence of an alternative pathway potentially mediated by the dorsolateral prefrontal cortex.

In the subgroup regression analysis, we observed significant differences in the relationship between MDS-UPDRS-III scores and cerebellar glucose metabolism across different PD subgroups. Specifically, a positive correlation was found in the lower LEDD group, shorter disease duration group, younger age group, and male patient group. In contrast, no correlation was observed in the higher LEDD group, longer disease duration group, older age group, and female patient group. These results suggest that in patients with shorter disease duration, lower LEDD, or younger age, the increased glucose metabolism in the cerebellum may play a compensatory role in the exacerbation of motor symptoms. Another noteworthy finding is the lack of a significant correlation in female patients. Preclinical evidence and small intervention studies suggest that estrogen may have a protective effect against dopaminergic damage19,20. Thus, estrogen may compensate for the cerebellar contribution to motor symptoms, and this hypothesis warrants further exploration.

As precise behavioral assessments and detailed mapping of cerebellar functional domains have advanced, it has become increasingly clear that a substantial portion of the cerebellum, particularly the posterior lobes, is involved in cognitive functions. Supporting this, fMRI studies have shown that cerebellar lobules VI, VII, and VIII are actively engaged in cognitive processes in healthy individuals21,22,23,24. Our study revealed that in PD patients, glucose metabolism in cerebellar lobules VI and VIII is associated with overall cognitive function, consistent with findings in healthy individuals.

Moreover, comparisons among PD patients with different cognitive statuses showed that glucose metabolism in cerebellar lobules VI and VIII was significantly higher in the PDD group than in the PD-MCI and PD-NC groups. The PD-MCI group also showed significantly higher glucose metabolism than the PD-NC group. To reduce the impact of confounding factors, we established regression models, which yielded results consistent with these observations.

In the subgroup regression analysis, we observed that the relationship between MOCA scores and cerebellar glucose metabolism showed smaller variations across different subgroups compared to the MDS-UPDRS-III scores. In the subgroup with lower education levels, a significant negative correlation between cerebellar glucose metabolism and MOCA scores was observed, whereas no such correlation was found in individuals with higher education levels. This result may be related to the stronger cognitive reserve in individuals with higher education levels. According to the cognitive reserve theory, higher education can help buffer against brain function decline25,26, potentially weakening the dependence between cerebellar metabolism and cognitive function. Significant correlations were observed in patients with shorter education years, as well as in all subgroups based on sex, age, and LEDD. Therefore, we hypothesize that the relationship between cerebellar metabolism and cognitive function may be stable. In PD patients, cerebellar involvement in cognitive function may be stronger than in motor function and may not be significantly affected by sex, age, or LEDD. Additionally, we observed a significant negative correlation between glucose metabolism in the cerebellar lobule VI and the MOCA score in the subgroup with a shorter disease duration. A similar trend was noted for cerebellar lobule VIII, although it did not reach statistical significance (p = 0.07). We hypothesize that cerebellar lobule VI may play a role in the early stages of the disease, contributing to cognitive impairment, while cerebellar lobule VIII could become more involved as the disease progresses to later stages. Further studies are needed to validate this hypothesis.

Building on these findings, we further analyzed the relationship between cerebellar metabolism and cerebral cortical regions. We found that the metabolism of lobules VI and VIII correlates positively with the metabolism of cerebral cortical regions, including the prefrontal, temporal, and occipital cortices. These brain areas are involved in various cognitive domains such as attention, executive functions, and memory27,28,29. This suggests that cerebellar clusters related to cognition may have close functional connections with these cortical areas, indicating that the cerebellum may help maintain cognitive functions or slow cognitive decline through interactions with supratentorial structures. This further supports the idea that increased cerebellar metabolism in PD patients may be a compensatory mechanism, helping to maintain relatively good cognitive function by coordinating metabolic activity with other brain regions.

In addition to the findings related to the cerebral cortex, our study also discovered that glucose metabolism in the cerebellar dentate nucleus is negatively correlated with cognitive function scores in PD patients. Both animal and human studies have demonstrated that the dentate nucleus exhibits distinct connectivity patterns: the dorsolateral region is associated with motor functions, while the ventrolateral region is linked to non-motor functions30,31,32,33. Research on neurodegenerative diseases has further highlighted the critical role of the dentate nucleus across different conditions. Specifically, in Alzheimer’s disease, cognitive function is closely associated with both functional connectivity (FC)34 and iron deposition35 in the dentate nucleus. In multiple system atrophy, both motor and cognitive functions are significantly related to the dentate nucleus’ FC36. In PD, prior studies have identified a correlation between iron deposition in the dentate nucleus and motor symptoms37. Our study additionally revealed a negative correlation between glucose metabolism in the dentate nucleus and cognitive function in PD patients, suggesting that the dentate nucleus may play a compensatory role in cognitive function in PD patients.

As Wu et al. proposed, the cerebellum plays a dual role in PD, involving both pathological and compensatory mechanisms38. However, how these two effects evolve throughout the progression of the disease remains unclear. It is likely that, although pathological changes are already present in the early stages of the disease, the compensatory effects may be enhanced during this phase. As pathological damage becomes more severe in the advanced stages, the compensatory effects may weaken or even eventually fail. This is consistent with our subgroup analysis results: in patients with shorter disease duration, an increase in UPDRS-III score was associated with enhanced glucose metabolism in the cerebellum. In contrast, in patients with longer disease duration, no such correlation between UPDRS-III score and cerebellar glucose metabolism was found. These findings align with the known functional anatomy of the cerebellum, which is connected to the cerebral cortex through the “cerebellar-basal ganglia-cortical circuit”39,40. This circuit plays a key role in motor control, cognitive functions, and emotional regulation. Recent fMRI studies highlight the cerebellum’s compensatory role within this network6,14, suggesting that its activity may help counteract dysfunction in PD. These compensatory mechanisms offer new insights and open potential avenues for therapeutic targets in PD treatment41,42.

However, this study has several limitations. First, in the PET statistics, the correlation results between the cerebellum and motor functions, as well as between the cerebellar motor lobules in the TD group and supratentorial structures, were not corrected for FWE, which may compromise the reliability of the findings. Second, due to the retrospective nature of this study, we cannot definitively conclude whether the observed cerebellar metabolic patterns in PD patients reflect compensatory mechanisms or pathological alterations. To address this, future longitudinal studies with therapeutic interventions are necessary to provide clearer insights. Lastly, the single-center design, relatively small sample size, and lack of biological and genetic data limit the external validity and generalizability of the conclusions.

In conclusion, our research confirms that the cerebellum has distinct functional subdivisions. The glucose metabolism in different cerebellar regions is positively correlated with the severity of motor and cognitive impairments in PD patients. We propose that this may represent a compensatory mechanism, suggesting that the cerebellum could be a potential therapeutic target for PD.

Methods

Subjects

In this retrospective study, we enrolled 104 PD patients treated at the China-Japan Union Hospital of Jilin University between September 2019 and May 2024. The diagnosis of PD was based on the 2015 Movement Disorder Society diagnostic criteria. For patients with a disease duration of less than five years, the diagnosis was confirmed by two senior movement disorder specialists through clinical evaluation and results from DAT-PET, FDG-PET, and MRI. Inclusion criteria were: confirmed diagnosis of PD, age between 50 and 85 years, and ability to complete cognitive and motor function tests. Exclusion criteria included: secondary Parkinsonism, Parkinson’s plus syndrome, severe psychological or psychiatric disorders, flawed imaging, or inability to comply with the assessments. Demographic data collected included gender, age, disease duration, education level, and LEDD. Motor symptoms were evaluated using the MDS-UPDRS-III. Cognitive function was assessed using the MoCA scale. In the motor domain, the TD subtype and the PIGD subtype are distinguished by calculating the ratio of the average tremor score to the average PIGD score, using the MDS-UPDRS-III, consistent with previous studies43. In the cognitive domain, patients were classified into three groups based on MoCA scores: those scoring below 21 were classified as PDD, those scoring between 21 and 25 as PD-MCI, and those scoring between 26 and 30 as PD-NC44,45. Patients who do not meet the criteria for these two subtypes are classified as the Intermediate PD group. This study was approved by the Ethical Review Committee of China–Japan Union Hospital, Jilin University (No. 2023092701). Being a retrospective analysis, all patient data were anonymized before analysis to ensure confidentiality. The committee granted a waiver of informed consent in compliance with ethical standards for retrospective studies.

PET image acquisition

After confirming the inclusion of PD patients, we retrieved their previously acquired PET imaging data to analyze the patterns of cerebellar glucose metabolism. The detailed PET imaging acquisition procedure is described below. The 18F-FDG injection was prepared by the Department of Nuclear Medicine, China-Japan Union Hospital, Jilin University (18F was produced by the HM-10 cyclotron of Sumitomo, Japan, and 18F-FDG prepared by automated synthesis module from Beijing PET Biotechnology Co., China, with radiochemical purity >95%). To ensure standardized conditions for the 18F -FDG PET/MRI acquisition, all participants underwent a fasting period of at least 6 h and were withdrawn from anti-PD medications for a minimum of 12 h prior to the procedure. Subjects were injected intravenously with 18F-FDG (3.7 MBq/kg), required to rest for 60 min in a quiet, light-protected environment, and then underwent a scan using UMI510 (United-Imaging, Shanghai, China). CT imaging was scanned from the calvarium to the basis crania at 120 kV, 240 mA current, 1.2 mm layer thickness and 10 s scan time. PET acquisition was conducted in 3D mode in the same field of view in 256 × 256 matrix with an axial field of view measuring 700 mm. Attenuation correction was performed with CT data, and reconstruction was performed with the ordered subset expectation maximization (OSEM). Coronal, sagittal, and transverse section images of the brain were obtained.

PET image transformation

FDG-PET images were preprocessed using Statistical Parametric Mapping 12 (SPM12; Functional Imaging Laboratory, UCL Queen Square Institute of Neurology, London, UK) on MATLAB R2022a (MathWorks, Sherborn, MA). The preprocessing steps included spatial normalization of each participant’s images to the Montreal Neurological Institute (MNI) brain template for subsequent statistical analysis, with all parameters selected according to the guidelines provided in the SPM manual. Following this, a Gaussian algorithm was applied for smoothing with a full width at half maximum (FWHM) of 4 mm to reduce noise.

Demographic analysis

The demographic and clinical characteristics were analyzed using R software (R Foundation for Statistical Computing, Vienna, Austria, R version 4.4.1, 2024-06-14). Gender distribution was analyzed using the χ2 test. For continuous clinical characteristics like age, disease duration, education level, LEDD, MDS-UPDRS-III scores, and MoCA scores, ANOVA was used to compare differences among cognitive status groups, with Bonferroni adjustment for post-hoc comparisons. For motor subtypes, differences between the two groups were compared using the Mann–Whitney test. A p-value of <0.05 was considered statistically significant.

PET statistics

The PET image data were analyzed using SPM12 in Matlab R2022a. We established multiple regression models with SPM12 to examine the relationships between patients’ motor function (MDS-UPDRS-III scores), cognitive function (MoCA scores), and cerebellar metabolism. Both models incorporated age, gender, disease duration, and LEDD as covariates to minimize potential confounding factors. The motor function model additionally included the MoCA scores as covariates, while the cognitive function model accounted for education level and MDS-UPDRS-III scores. Initially, we focused on regions that were significant at a threshold of p = 0.005, corrected for FWE. In the absence of significant cerebellar regions at this threshold, we relaxed the criteria by removing FWE correction, while maintaining the p = 0.005 threshold and considering only cerebellar regions with more than 30 voxels. Finally, significant cerebellar regions were visualized using MRIcroGL software (v1.2.20210317) on the cerebellar SUIT template (v3.7)46.

Next, we used the Marsbar toolbox to extract the mean glucose metabolism values from the significant cerebellar cortical regions. These values were normalized using the formula: normalized metabolism value = (metabolism value/whole-brain metabolism value)× 50 × 100%47. To visually illustrate the metabolic differences among different PD subtypes, we created violin plots that showcased the results of ANOVA (with post hoc comparisons adjusted using the Bonferroni correction) and Mann–Whitney tests. To balance the confounding factors in between-group comparisons, we conducted ANCOVA analysis. Additionally, to represent the correlation and trends between glucose metabolism and motor and cognitive scores, we plotted regression curves.

Regression analysis and subgroup analysis

The regression models and subgroup analyses were conducted using R software.

To quantify the effects of motor and cognitive function on glucose metabolism in key cerebellar regions while controlling for potential confounding factors, we constructed regression models. The normalized glucose metabolism values served as the dependent variable (Y), with the MDS-UPDRS-III score or MoCA score as the primary independent variable (X). The covariates were consistent with those used in the PET image analysis.

To further investigate the variations in the relationship between cerebellar glucose metabolism and motor/cognitive functions across different subgroups, subgroup analysis was conducted in this study. The subgrouping factors included age, sex, disease duration, education level, and LEDD, with sex treated as a categorical variable. For other continuous variables, median splits were applied to ensure balanced sample sizes across the groups. The construction of the regression models followed the same procedures as described previously.

Correlation analysis with supratentorial structures

Finally, to further explore the association patterns between significant cerebellar regions and supratentorial structures, and to identify potential links between cerebellar glucose metabolism and supratentorial metabolic activity, we performed regression analysis on the extracted glucose metabolism values and FDG-PET images. Initially, we focused on regions that were significant at a threshold of p = 0.001, corrected for FWE. In the absence of significant cerebellar regions at this threshold, we relaxed the criteria by removing FWE correction, while maintaining the p = 0.005 threshold and considering only regions with more than 40 voxels. Statistical results were visualized on standard brain MRI T1-weighted images using MRIcroGL software, providing a comprehensive view of these association patterns.